PASS Flashcards
What are the 2 types of statistics
Descriptive , inferential (analysed)
What is evidence based medicine
Conscientious, explicit and judicious use of current best evidence in making decisions about care of individual patients
What is epidemiology
Study of distribution and determinants of health-related states or events in specified populations and application to health problems
Characteristics of surveillance and descriptive studies
Studies distribution
One group studied, no explicit hypothesis, development of possible hypothesis
Analytical studies
Study determinants
2 or more groups
Definite hypothesis
Reject or accept
Experimental studies are always
Analytical
2 types of observational study
Descriptive and analytical
2 types of descriptive study
Ecological studies and cross-sectional surveys
2 types of analytical studies
Case-control
Cohort
Requirements of sample population
Representative, unbiased, precise
2 types of validity
Internal and external
What is internal validity
Freedom from confounding, bias or random error
What is external validity
Degree to which conclusions can be applied to the population of interest
2 types of error
Chance or bias
Why do chance errors happen
Due to sampling variation, reduces as sample size increases
2 types of bias
Selection bias or information bias
Reasons for selection bias
Study sample not representative
Group selection within study not comparable
Healthy worker effect
Information bias examples
Recall error
Observer/interviewer error
measurement error
Misclassification
What is prevalence
Absolute risk
Proportion of people with a disease
What is incidence
Absolute risk
Number of new cases within a given time frame
What is incidence rate ratio
Compares incidence rate in 2 groups
IR1/IR2 = IRR
What is odds ratio
Comparison of odds of disease in one group compared to another
Ratio of ratios
what is risk difference
Absolute risk of A - Absolute risk of B
No difference = 0
What are person years
Sum of total time of everybody followed up in study
People x years
What is 95% confidence interval
Range within which we can be 95% certain that the true value lies
Wider 95% confidence interval if
Greater variation in population values
Smaller sample size
How to calculate 95% confidence interval
Error factor = e to the power of 2 x square root of 1/a + 1/b
What 95% confidence ratio suggests findings are not significant
If it spans over 1
Upper and lower boundary calculations
IR x ef = upper
IR / ef = lower
Upper-lower = x
What happens if 95% confidence interval spans over 1 e.g. 0.5-8
Fail to reject the null hypothesis and no statistical significance
Issue when comparing groups
Confounding
Confounding variable must influence
Both the group and the thing being tested
Solutions to confounding
Match important confounders
Weighted average
Standardised mortality ratio
Ecological studies key points
Identify groups of people to study (not individuals)
Data on group-level characteristics
Observational
Issues with ecological studies
Measurement variation Confounding Chance (random error)
Cross-sectional survey key points
Survey/series of surveys
Exposure and outcome measured simultaneously
Determines prevalence mainly
Analysis of individuals
Example of ecological study
Colon cancer incidence per 100,000 women and per capita daily meat consumption
Example of cross-sectional survey
Effect of aircraft noise exposure on heart rate during sleep in population living near airports
Issues with cross-sectional survey
Sampling bias
Responder/participant bias
Chance (random error)
Confounding
Advantages of cross-sectional survey
Cheap
Fast
Reflective of real life
Case control study key features
Always retrospective Identify group of cases and non-cases (controls) Ascertain previous exposure status Compare level of exposure in each group Analyse odds ratio
What is a nested case-control study
Collection of data from evolving outcome and exposure database of a concurrent or prospective cohort study
Advantages of nested case control study
Incidence rates calculated
Population for sampling already defined
Can collect more detailed information for a minority of participants
Advantages for case control study
Good for rare diseases
Cheap
Quick
Can study multiple exposures for a single outcome
Issues with case control study
Selection bias
Information bias (misclassification)
-Non-differentiated (randomly inaccurate measurement)
-Differentiated (systematic, recall bias, assessor bias, data collection errors)
Confounding
Chance (random error)
What is a cohort study
Always prospective
Group individuals according to level of exposure
Select outcome free individuals
Ascertain outcomes for everyone
Compare incidence rates for each exposure group
Analysis of cohort study
Odds ratio/rate ratio
Comparisons externally e.g. standardised mortality ratio or internally e.g. IRR
Advantages of cohort study
Enable derailed and prospective assessment of exposure, outcomes and confounders
Studying a range of different outcomes, rare exposure, whether exposure precedes outcome, conditions that fluctuate with age
Issues in cohort study
Loss to follow up - differential loss, survivor bias Information bias Confounding Chance (random error) Expensive Take long time Large and resource intensive
Example of cohort study
5000 people followed up from age 55 for 10 years
2000 smokers —> 200 developed lung cancer
3000 non smokers —-> 20 developed lung cancer
Describing a study
Study design- PICO Population Intervention/exposure Comparison/control Outcome
What is SMR
Standardised mortality ratio
SMR equation
Observed number of deaths/ expected number of deaths
What bias is always present
Sampling bias
Random error
Also somewhat confounding
In what study is confounding highest
Ecological
